关键词: 波段选择,子空间划分,Bhattacharyya距离,灰色关联决策 中图法分类号TP751文献标识码A

Abstract: As researchers’ demand for the quality of the spectral information in hyperspectral images gradually increases,the characteristics of hyperspectral images impedes the further information extraction to the images.The existing single band selection method can not fully consider the criterias about "information content,correlativity,class separability",and the results are inevitably restricted by other index measurements．Using the quality of grey system theory and taking the small sample,small information and uncertainty system as research subjects can do the grey incidence decision on the basis of subspace partition,overcoming the independence and incompatibility of the single-index measure.Therefore,aiming at the growing demand about ensuring the separability of conjugate class,this paper put forward a band selection method which can synthetically consider the results of other single band selection mehtods with grey incidence decision.Finally,an experiment was made and it was compared with common fusion methods.

Key words: Band selection,Subspace partition,Bhattacharyya distance,Grey incidence decision